A tool for scraping tweet ids from the Twitter website.
Tweets collected from Twitter's APIs provide metadata about the tweets that is machine-readable (in JSON) and may not be available from the website. For the purposes of archiving and/or analyzing tweets this metadata is potentially significant.
Unfortunately, Twitter's APIs don't support getting a comprehensive set of a user's tweets. Twitter's statuses/user_timeline REST API method only allows collecting the last 3,200 tweets. Similarly, Twitter's Search API will only provide tweets from the last 6-9 days.
twitter-scraper attempts to support getting a comprehensive set of a user's tweets (with optional date constraints). It accomplishes this by making requests to Twitter's website search (which is different than the Search API) and extracting tweet ids. These tweet ids can then be passed to twarc to retrieve from Twitter's REST API (aka "hydrating").
-
Install Python, Pip, Git, and Chrome.
-
Clone this repo:
git clone https://github.com/justinlittman/twitter-scraper.git
-
Install ChromeDriver. On a Mac, this can be done with
brew install chromedriver
. On Ubuntu:wget http://chromedriver.storage.googleapis.com/2.24/chromedriver_linux64.zip unzip chromedriver_linux64.zip sudo mv chromedriver /usr/bin/
-
Install Selenium:
pip install selenium
-
Install Twarc:
pip install twarc
python twitter_scraper.py -h
usage: twitter_scraper.py [-h] [--since SINCE] [--until UNTIL]
[--exclude-retweets] [--delta-days DELTA_DAYS]
[--wait-secs WAIT_SECS] [--debug]
screen_name
positional arguments:
screen_name
optional arguments:
-h, --help show this help message and exit
--since SINCE Tweets since this date. Default is 2011-04-05.
--until UNTIL Tweets until this date. Default is today.
--exclude-retweets
--delta-days DELTA_DAYS
Number of days to include in each search.
--wait-secs WAIT_SECS
Number of seconds to wait between each scroll.
--debug
To collect @realDonaldTrump's tweets between January 1, 2016 and April 1, 2016:
-
Run twitter_scraper and write the tweet ids to a file.
python twitter_scraper.py @realDonaldTrump --since=2016-01-01 --until=2016-04-01 > tweet_ids.txt
Leave your system alone while twitter_scraper is running. I received inconsistent results while I was doing other work on my system while twitter_scraper was running. Better yet, use a VM.
Tip: You can get the date that a user joined Twitter from the user's account page.
Tip: Timestamp the tweet id file by using
> tweet_ids_$(date -d "today" +"%Y%m%d%H%M").txt
Tip: To account for the inconsistency, you may want to run multiple scrapes. They can be merged with:
cat tweet_ids_*.txt | sort -u > tweet_ids.txt
-
Hydrate the tweet ids with Twarc and write to a file. You will need to provide Twarc with a set of Twitter API keys. For more information, see Twarc's documentation.
twarc.py --hydrate tweet_ids.txt > tweets.json
This work was inspired by the Trump Twitter Archive.
And appreciative (once again) for Ed Summer's twarc.